Remove spurious spaces at EOL

Discard 0.9.3-compat branch

"or" for "if" in getmodulename; shorter comment on backwards compatibiliy

Added tag hgkw_standalone_setup for changeset 99dc49c5bcfb

Setup standalone module install, version tracking

Make all ui config and kwfmatcher a local repo attribute See Mercurial changeset: 5d9ede002453

No reposetup when no filenames are specified for expansion Second stab at stopping reposetup. This time kwtemplater is still initialized in kwfilelog and pretxnkw, and not in reposetup. Watchout: Giving ui an extra-attribute in reposetup might still be problematic.

Add test

Add licence (again)

Put .hg* files in exclusion list at declaration

Prepare test: split doc into help and comment Minor cleanups in function docs

Be silent if no filenames are configured for expansion The debug message would only be visible on commit anyway.

Kill buggy branch

FIX: no appending of attributes to repo, back to b2cc6a8d4a18 The idea was to good/unclean to be true. Raised unknown exception on commit when exiting $EDITOR with return 1. Integrate "innocent" changes since then though.

Describe substitution mechanism in doc

Debug info, not warning, when no files configured for expansion

Cycle filename patterns only once

Get modulename via function; set kwrepo attribs via direct reference

Move all that can be done only once per repo into reposetup Actually kwrepo is not set up if there aren't any files configured for keyword substitution. Stuff that now is done at reposetup and not at every filelog init or hook: 1) filename matching function 2) compilation of keyword regex 3) templates and changeset templater kwtemplater as an appended class should prevent namespace conflicts.

Extend templater.common_filters in kwtemplater

Document reposetup; recommend loading on per-repo basis

Implement configurable exclusion of files from expansion

Only initialize kwtemplater in kwfilelog if needed

Use full power of util.matcher for kwfmatches()

Simply dict() keywordmaps configitems

Emergency brake against multiline expansion As expansion always happens ad hoc, is not tracked, multiline or incremental expansion do not seem worth the trouble. If at all, it would make sense with static, hardcoded keywords.

More doc, link to KeywordPlan

No need to read changelog in show(); node mandatory arg for expand()

Just re-bin the node kwarg of the hook, instead of using changelog

Same as for hardcoded branch: omit hooktype test, return w/o explicit value

Omit hook test; return w/o explicit value (==False)

No clumsy setting of node in kwfilelog.add, pass as kwarg

Expansion entirely in kwfilectx.expand()

Implement custom keyword map templates in hgrc More expensive than hardcoded keywords. But, of course, more flexible. ATM templates can only be set in hgrc, and perhaps this is enough.

Indentation cosmetics

Clean keyword arg assignment in kwfilelog.add; rename utc() to utcdate()

Init context.filectx only once per file with class kwfilectx This way context.filectx is not initialized for every keyword match.

Added tag very_cvs for changeset 2e930f842242

Very CVS-like format

Functions instead of variables in kwexpand() Should be slightly faster as filectx is asked only on demand.

Module detection cleaner and more reliable Stolen from mercurial.lsprof

Only extend sys.path if needed

Discard head of modular branch

Discard head of decodefilter branch

Make pretxncommit hook work even if extension not in hgext folder

Cosmetics: var name kwfmatchers, remove some line continuations

Add keyword; comment backwards compatible import

Implement self initializing pretxncommit hook This is more expensive than overriding commit(), but a) uses more "official" interface b) less code c) easier to maintain Get all changes up to cee5fef33cf8 in here as well.

Fix timezone expando in doc

Make changes from last cset 474b415433a1 available to 0.9.3

Unexpanded storage hopefully covered now by adding kwfilelog.add (again!) All filelog methods that act on text/data should be consistent. Principle: read expanded kw's; unexpanded kw's for "internal" workings. kwfilelog.iskwcandidate() to check whether keyword action required. More doc.

Check against keyword config in kwfilelog.cmp; be silent in kwfilelog (0.9.3)

Check against keyword config in kwfilelog.cmp too; be silent in kwfilelog

Keyword myself 0.9.3

Keyword myself

Branch for compatibility with hg-0.9.3 release See Message-ID: <91340540@bs1.sp34.ru>.

Fix indent in entire kwrepo.commit function

No keyword substitution in cmp(); no overriding of size() ATM This hopefully solves the merge/no conflict issue brought up by Boris Samorodov in Message-ID: <12807831@srv.sem.ipt.ru>.

Added tag pure_extension for changeset 4c5d9635b517

Pure extension by inserting keyword expansion localrepo.localrepository.commit Replaced self.ui with ui in commit(). This works, or seems to work, but can it be done without a complete copy of commit()?

Remove copyright for the moment

Remove kwfilelog.add() -- seems to work with current kw_re

Make kwfilelog.add respect filename config

Unify obtaining filename matchers; only debug messages

Imports specific to functions

Omit 1 implicit if-clause

Added tag r0.1 for changeset 1fe48bf82d05

Remove more internal comments

List supported keywords in doc

Only 1 all-purpose regex, compiled at load

lambda m: instead of kwexpander(matchobj)

Extract all information in kwexpand()

Move context.filectx into if-clause where it's actually needed

Use localrepository method to obtain full path

Cleanup comments; refer to hybridity of script in doc

Change location of date format comment

Make inheritance clearer in size() and add()

Reintroduce simpler kwfilelog.add(); simplify kwfilelog.read() Without add() strange things happen to changectx().

Use self._path directly instead of assigning f to context.path()

Override filelog's size(), cmp() as well; remove add() Try whether we /really/ don't need add(). size() and cmp() are needed to avoid eg: $ hg st $ touch file $ hg st M file

add header and start documentation

Branch standalone extension, including pretxncommit hook function Slower, but easier to integrate.

Make keyword.py depend on resurrected kwutil.py This makes pretxnkw perhaps a bit slower, but is more readable.

Exclude import command too

Exclude tag cmd from pretxcommit hook; exclude .hg files Return unchanged data in superclass read() after every if clause. Perhaps a bit more readable.

Added tag cvs_scheme for changeset ba000e29ecf3

Implement near CVS compability with more than one keyword Keywords are set as local vars in kwexpand() and then eval'd. TODO: (sticky) tag?

Make patterns stricter again, to avoid unvolontary expansions

Rename "text" to "data" always as set before call to util.binary()

Reflect new scheme in function docstring

Remove all hgkwencode, kwexpander.py and updatekw.py This scheme should work just with: extension keyword.py pretxncommit-hook pretxnkw.pretxnkw

Standalone version for keyword.py Check "expand" patterns in the [keyword] config section. Removed keyword in file itself to simplify patterns. Revert (again) to basename.

Change config opt "expand" to accept filename patterns Not nice yet; but passes tests together with pretxnkw. TODO: Find out what filelog.add() does exactly.

Interim test version in connection with keyword.py Simplify grep pattern. Outcomment filename lookup, will be changed to configitems('keywords'). Update stuff will be removed. And the whole thing will go into pretxncommit. kwupdate.py will be removed.

Make keyword do my hgkeyword; use basename again add() needs only str.replace(). TODO: Make even more compatible with cvs' $Id$? (date UTC, slashes) Check config before reading. And!: Use config file patterns, in cooperation with pretxnkw. This will expand keywords alread on commit!

Add original keyword extension by Thomas Arendsen Hain

Update function descriptions

Add TODO for walk back through file history on update

Test for binary

Clean up variable names; do not explicitly return False

Do not explicitly exit False; rename node1/2 to r1/2, p1/2 to n1/2

Do not explicitly return False at end

Reparse cmdline to avoid expansion in uncommitted files

Make wwritekw() and all vars local

Throw out demandimport -- at least temporarily

Double check encode filtered filenames

Added tag universal_scheme for changeset 536c1797202d

Implement $Hg$ scheme with update hook The pivotal line for update hook is: repo.dirstate.update(kwupdates, 'n') This forces hg to consider the freshly written files as not modified. Thanks to wfile(), this keeps executable bits etc. Still needs more testing. No need to check for basename/filename in keyword trigger. update hook does not need re. TODO: Walk back in history, if last change of file didn't happen in 1 of the provided changesets (update/merge?)?

Simplify for $Hg$ expansion scheme

Use the "parent1|2" kwargs to detect file state Can parent2 be of further use? Like in parsing back history? Adapt keyword trigger to $Hg$ scheme.

Simplify filter pattern for $Hg$ scheme Adapt keyword trigger as well.

Start universal $Hg$ keyword expansion

Rename parent1 to p

Rename lr variable to repo

Adapt to filename instead of basename style

Short "f" for "filename" iteration variable

Added tag filename for changeset 56a61a5c696d

Switch to complete filename in first keyword field This will hopefully allow to iterate over the stream in hgkwdecode, and lookup the relatively latest revision of corresponding file.

Use localrepository's wjoin() to obtain absolute path

Use wfile instead of wopener; read and write string not binary

FIX: range of status selection; copyfile using absolute path

Add pretxnkw module for updatehook branch

Add pretxnkw module for commits

updatehook branch: add kwexpander and updatekw modules WARNING: Do not use this branch for production! Update hook changes to working directory are detected by hg. The only /hook/ to escape hg's attention seems to be pretxcommit. And that probably only due to bug. Kept branch for archiving purposes.

decodefilter branch: add decode filter with kwutil module

Make pattern compilation local to function

Shorten variable and function names

Add simple [encode] filter in python The aim is to go for $Hg$ without basename. Perhaps this is feasable with update hooks &Co. No way to detect a filename while stream writing. Even though hg /knows/ the filenames, but does not provide an API or hook, unfortunately.